Download virus and its age-specificity in Japan

R a p i d c o m m u n i c a ti o n s
Transmission
p ot e n t i a l o f t h e n e w i n f l u e n z a
v i r u s a n d i t s a g e - s p e c i f i c i t y i n J a pa n
A(H1N1)
H Nishiura ([email protected])1, C Castillo-Chavez2, M Safan3, G Chowell2
1.University of Utrecht, Utrecht, the Netherlands
2.Arizona State University, Tempe, Arizona, United States
3.Mansoura University, Mansoura, Eg ypt
On 16 May 2009, Japan confirmed its first three cases of new
influenza A(H1N1) virus infection without a history of overseas
travel, and by 1 June, 361 cases, owing to indigenous secondary
transmission, have been confirmed. Of these, 287 cases (79.5%)
were teenagers (i.e. between 10 and 19 years of age). The
reproduction number is estimated at 2.3 (95% confidence interval:
2.0, 2.6). The average number of secondary transmissions involving
minors (those under 20 years of age) traced back to infected minors
is estimated at 2.8. That is, minors can sustain transmission even
in the absence of adults. Estimates of the effective reproduction
number Rt moved below 1 by 17 May. Active surveillance and
public health interventions, including school closures most likely
have contributed to keeping Rt below one.
Introduction
The reproduction number R, the average number of secondary
cases generated by a single primary case, of the new influenza
A(H1N1) virus, is a key quantitative measure for assessing
pandemic potential [1]. In the ongoing epidemic of the new
influenza A(H1N1) virus, early studies suggested that R ranged
from 1.4-1.6 [2] and some estimated it to be as high as 2.2-3.1
[3]. Estimates in 1.4-1.6 range for the new influenza A(H1N1)
virus are lower than estimates based on data from, for example,
the fall wave of the 1918 influenza pandemic [4,5]. The present
study investigates indigenous secondary transmissions of the new
influenza A(H1N1) virus in Japan, not only estimating R but also
exploring its age-specificity.
worker at Tokyo-Narita airport) [6]. Figure 1 shows the geographic
distribution of 361 indigenous cases. Cases outside Osaka and
Hyogo prefectures had travel histories to Osaka or Hyogo before
their illness onset. The index case(s) (who may have remained
asymptomatic [7]), with a history of overseas travel, has (have)
yet to be identified. Furthermore, there are no known cases prior
to the five confirmed cases that developed the disease on 9 May
in Hyogo (Figure 2A). The triggering event may be associated with
Japan’s two-week festive break, the “golden week”, just before 9
May, when people may have travelled to and returned from Mexico,
United States and Canada.
Figure 1
Spatial distribution of the epidemic of new influenza
A(H1N1) virus infection in Japan. Cumulative number of
confirmed indigenous cases, as of 1 June 2009 (n = 361)
Methods
Epidemiological description of the epidemic
On 16 May 2009, three high school students in Kobe city, Hyogo
prefecture, without a history of overseas travel, were confirmed
as infected with the new influenza A(H1N1) virus. Confirmatory
diagnosis in Japan requires influenza-like symptoms and a laboratory
diagnosis which is made either by virus isolation, real-time PCR
or a significant increase in neutralising antibody titre against the
virus. Further confirmed diagnoses followed predominantly in
Hyogo and Osaka prefectures. The increased number of infections
among particular age groups was most evident in the data from
prefectures where most secondary cases were found among high
school students attending different schools.
By 1 June, the Ministry of Health, Labour and Welfare of Japan
had reported 371 confirmed cases, including nine imported cases
and one case traced back to a distant international airport (i.e. a
Note: Cases in Tokyo, Saitama, Shiga and Kyoto had travel history to either
Hyogo or Osaka prefecture before illness onset. Kobe city, where first
three cases were diagnosed, is a capital city of Hyogo prefecture.
E U R O S U R V E I L L A N C E Vol . 14 · I ss u e 22 · 4 J u n e 20 0 9 · w w w. e u ro s u rve i ll an c e . o rg 1
We analysed the temporal incidence distribution of confirmed
cases for this epidemic (Figure 2A). The known dates of illness
onset are used except for a fraction of the confirmed cases in Kobe
city (45; 12.5%) whose dates of onset have yet to be fully clarified.
Since the known median time from onset to diagnosis in Kobe
has been estimated at 1.0 day [8], it is assumed that the dates of
onset among the 45 cases in Kobe were 1 day before their date of
diagnosis. We observed that by the time the first three cases had
been confirmed (16 May), the epidemic curve was just about at its
peak. 16-17 May fell on a weekend, and all schools in Osaka and
Hyogo were officially closed for one week starting on 18 May. Figure
2B displays the age-distribution of the 361 confirmed cases, which
is concentrated in the teenage population. We see the age-specific
window (10-19 years of age) that includes 287 confirmed cases
(79.5%; 95% confidence interval (CI): 75.3, 83.7).
Epidemiological analysis
Taking into consideration the high levels of uncertainty related
to the invasion of a population by a novel influenza virus, three
different methods are used to estimate the transmission potential
of the new influenza A(H1N1) virus. To concentrate on the
transmission potential in Japan, all nine imported cases and one
case that is not associated with indigenous transmission in Hyogo
and Osaka were removed from the following analyses.
Model 1 (M1)
Estimation of R using the intrinsic growth rate [3,5]. The
intrinsic growth rate r, is estimated via a pure birth process [9].
The likelihood is proportional to:
(
)
exp − r ∑ i = 0 C ( i ) (1 − exp ( − r ) )
t −1
C ( t ) − C ( 0)
where C(t) denotes the cumulative number of cases on day
t. C(0) = 5 and t = 0 represents 9 May. The generation time (GT) is
assumed to follow a gamma distribution with mean μ= 1.9 days and
coefficient of variation ν = 47% [2]. R is subsequently estimated
using the estimator [10]:
1
2 v2
(1 + r µ v )
Given that many serial intervals reported from Spain are longer
than 1.9 days [7], the uncertainties surrounding GT estimates are
partially addressed through a sensitivity analysis of R to variations
in the mean GT in the range from 1.3-4.0 days. The exponential
growth phase is assumed to have a mean duration of 8 days but
windows in the 8±2 days were also used.
Model 2 (M2)
The effective reproduction number Rt, the average number of
secondary cases generated by a primary case at time t, is estimated.
The daily growth rate rt is used to estimate Rt following the approach
described elsewhere [11]; the distribution of GT and the estimator
of R used are the same as those used in M1. The mean GT is
assumed to be 1.9 days but varying in the 1.3 to 2.5 days range [2].
Figure 2
Time- and age-specificity of the epidemic of new influenza A(H1N1) virus infection in Japan
A) Epidemic curve of confirmed indigenous cases according to the date of illness onset, as of 1 June 2009 (n = 361)
B) Age distribution of confirmed indigenous cases, as of 1 June 2009 (n=361)
A
B
C
360
20
180
120
10
60
0
0
27
10-19
0-9
29-May
27-May
25-May
23-May
21-May
19-May
17-May
15-May
13-May
11-May
9-May
7-May
15
11
13
5
3
60 and
older
30
240
50-59
40
40-49
Confirmed cases
50
287
300
30-39
Age group
20 and older
0-19
60
Confirmed cases
B
20-29
A
70
Age (years)
Note: None of the confirmed cases had recent history of overseas travel (except for one case in Wakayama). The nine cases, believed to have become
infected abroad, and one case, arising in a worker at Tokyo-Narita airport, are excluded from these figures.
The dates of illness onset for each confirmed case are reported by prefectural governments, except for a fraction of cases in Kobe city where cases with
unknown dates of onset are assumed to have developed the disease one day before the confirmatory diagnosis (based on published median estimate [8]).
It should be noted that the dates of onset are based on preliminary reports and have yet to be refined.
Arrow A indicates the date on which the first three cases were diagnosed in Kobe city. All schools in Hyogo and Osaka were closed between the dates
signalled by the arrows B and C.
2
E U ROSU R V E I L L A N C E Vol . 14 · I ss u e 22 · 4 J u n e 20 0 9 · w w w. e u ro s u rve i ll an c e . o rg
Model 3 (M3)
The role of age-specificity in transmission is analysed using
estimates of the next-generation matrix, K (Figure 3). First, we
aggregate the population in two age groups, minors and adults.
Second, since the mean GT is approximately 2 days [2], the daily
number of cases during the exponential growth phase (i.e. first
8 days) uses as its unit of time, two-day intervals (i.e. cases, c,
in days 1 & 2, 3 & 4, 5& 6 and 7 & 8 are grouped). Third, the
expected value of cases in age-group i of grouped-generation τ,
E(ci(τ)), is modelled by Riici(τ-1)+Rijcj(τ-1) (fort = 2, 3 and 4) where
Rgh is the element of K that corresponds to the average number
of secondary cases in group g caused by an infected individual in
group h. We estimate the entries in the matrices, assuming two
different mixing patterns modelled via two unknown parameters by
means of Poisson regression (Figure 3).
Results
The intrinsic growth rate r, is estimated at 0.47 (0.40, 0.56) per
day. Accordingly, M1 gives an R estimate of 2.3 (95% CI: 2.0, 2.6).
Figure 4A illustrates the sensitivity of R to variations in the mean
GT in the range 1.3-4.0 days. The corresponding R estimates lie
in the 1.8 to 4.8 range. Variations in the initial growth phase (i.e.
±2 days) do not greatly influence R; i.e. the expected values of R
lie in the 1.9 to 2.3 range. The exclusion of the less documented
cases in Kobe lead to an R estimate of 2.0 (95% CI: 1.7, 2.3).
Use of M2 suggests that Rt peaked on 14 May (Figure 4B). On
17 May, the day after a press release announced the first three
confirmed diagnoses, Rt declined below 1. Under active surveillance
efforts and school closures, Rt was kept below 1 thereafter.
Consistent temporal patterns of Rt are seen using different values,
except for slight increase and decrease in Rt estimates, for GT mean
values in the 1.3-2.5 day-range.
Figure 3
Next-generation matrix
Using M3, the next-generation matrix, K1 estimate, under the
separable mixing assumption is
 2.82 0.32 
Kˆ 1 = 

 0.32 0.04 
while our K2 estimate based on a qualitative assumption of
WAIFW (who acquired infection from whom) matrix is
 2.82 0.29 
Kˆ 2 = 

 0.29 0.29 
The host-specific reproduction number [12] for minor, i.e. the
average number of secondary minor cases generated by a single
primary minor case was 2.8 under K1 and K2. Hence a population
of minors can sustain the chains of secondary transmission even in
the absence of adults (i.e. for this epidemic “minors” are the “core”
group). Our estimate of R based on M3 is the largest eigenvalue
of K, and R is estimated at 2.9 for both matrices. These estimates
are slightly greater than R estimates based on M1; when the mean
and variance of GT is 2.0 days and 0 days2 (i.e. if GT is constant,
following a delta function), our R estimate is 2.6.
Discussion
Two important conclusions can be drawn from our epidemiological
analyses. Firstly, the reproduction number R of the new influenza
A(H1N1) virus in Japan is estimated to be as high as 2.3, a value
that is significantly higher than that recently reported [2]. The
pandemic potential of this virus in Japan may be higher in terms of
transmission potential than in other areas of the world. In particular,
it should be noted that our estimate of R is greater than published
estimates for seasonal influenza epidemics in temperate countries
[13]. Given that our R estimate has been tested for robustness to
uncertainty to mean GT, it seems plausible that high contact rates
among teenagers (when compared to other populations) may be one
of the main drivers of this epidemic. From a transient increase in
Rt around 14 May, our high estimate of R may reflect the existence
of few highly connected clusters of cases among “cliques” of high
school students. There may be additional contributing factors to
variations in our R estimates, including cross-protective immunity
due to previous exposure to other closely related influenza viruses.
Secondly, our age-specific estimates support the view that minors
can sustain transmission of the new influenza A(H1N1) virus among
themselves. Available data are not enough to investigate the precise
role of age-specific effects (e.g. different roles of transmission
among infants, primary-school, high-school and university students)
due to small case counts. Nevertheless, we believe that the
population of minors could play a key role as a “reservoir” for
sustained chains of secondary transmission, despite the fact that
cases in this group include those infected in some atypical school
clusters. Should further data confirm these results then the value
of public health interventions targeting minors (closing schools and
further contact restrictions between minors) could be effective in
controlling further outbreaks in Japan and other countries.
Note: Each element of the next-generation matrix, i.e., R cc , R ca , R ac and
R aa , denotes the average number of secondary transmissions caused by
a single primary case for child-to-child, adult-to-child, child-to-adult
and adult-to-adult transmissions, respectively (note that here “child”
represents “minor”, aged from 0 to 19 years). The reproduction number
R, for the whole population, is given by the largest eigenvalue of the
next-generation matrix. By making qualitative assumptions A and B, two
parameters, a and b, are estimated.
Our estimates of Rt provide a quantitative measure of the timeevolution of the “force” of the epidemic. Although the dates of
onset have yet to be refined and, thus, the precision of Rt estimate
may have been influenced by possible delay in diagnosis and
reporting, Rt declined below 1 one day after the news of the first
E U R O S U R V E I L L A N C E Vol . 14 · I ss u e 22 · 4 J u n e 20 0 9 · w w w. e u ro s u rve i ll an c e . o rg 3
three confirmed diagnoses. Thereafter, the implementation of active
surveillance programmes, including contact tracing, combined with
school closures, most likely have contributed to keeping Rt below 1.
R is useful for assessing transmission potential, and it is one of
the ways of assessing pandemic potential. This study puts emphasis
on quantifying the impact of contact patterns on the transmission
potential, factors that vary across space and time. Thus, further
analyses of R for the new influenza A(H1N1) virus in different
settings are needed to better quantify the role of uncertainty
and heterogeneous patterns of transmission in these estimates.
Validation of our quantitative understanding of the role of agespecific transmission should lead to improved effectiveness of
age-specific control measures.
Acknowledgements
The work of H Nishiura was partly supported by The Netherlands
Organization for Scientific Research (NWO; grant ID: 851.40.074).
Refe re nces
1. Coulombier D, Giesecke J. Why are Mexican data important?. Euro Surveill.
2009;14(19):pii=19212. Available from: http://www.eurosurveillance.org/
ViewArticle.aspx?ArticleId=19212
2. Fraser C, Donnelly CA, Cauchemez S, Hanage WP, van Kerkhove MD, Hollingsworth
TD, et al. Pandemic potential of a strain of influenza A (H1N1): early findings.
Published 11 May 2009 on Science Express. Doi: 10.1126/science.1176062.
Available from: http://www.sciencemag.org/cgi/content/abstract/1176062
3. Boëlle PY, Bernillon P, Desenclos JC. A preliminary estimation of the
reproduction ratio for new influenza A(H1N1) from the outbreak in Mexico,
March-April 2009. Euro Surveill. 2009; 14(19): p11=19205. Available from: http://
www.eurosurveillance.org/ViewArticle.aspx?ArticleId=19205
5. Andreasen V, Viboud C, Simonsen L. Epidemiologic characterization of the 1918
influenza pandemic summer wave in Copenhagen: implications for pandemic
control strategies. J Infect Dis. 2008;197(2):270-8.
6. Ministry of Health, Labor and Welfare, Japan. Influenza A(H1N1). Tokyo: Ministry
of Health, Labor and Welfare; 1 June 2009. Available from: http://www.mhlw.
go.jp/english/topics/influenza_a/index.html
7.
Surveillance Group for New Influenza A(H1N1) Virus Investigation and Control
in Spain. New influenza A(H1N1) virus infections in Spain, April-May 2009. Euro
Surveill . 2009; 14(19): p11=19209. Available from: http://www.eurosurveillance.
org/ViewArticle.aspx?ArticleId=19209
8. National Institute of Infectious Diseases, Kobe Institute of Health. Interim
report on clinical presentation of the novel influenza A(H1N1) cases reported
from Kobe City. Tokyo: National Institute of Infectious Diseases; 19 May 2009.
Available from: http://idsc.nih.go.jp/disease/swine_influenza_e/idsc_e2009/
clinical_epi_kobe.html
9. Bailey NTJ. The elements of stochastic processes with applications to the
natural sciences. New York: Wiley; 1964.
10. Roberts MG, Heesterbeek JA. Model-consistent estimation of the basic
reproduction number from the incidence of an emerging infection. J Math
Biol. 2007;55(5-6):803-16.
11. Nishiura H, Chowell G, Heesterbeek H, Wallinga J. The ideal reporting interval
for an epidemic to objectively interpret the epidemiological time course. J
R Soc Interface. 2009. [In press]
12. Nishiura H, Hoye B, Klaassen M, Bauer S, Heesterbeek H. How to find natural
reservoir hosts from endemic prevalence in a multi-host population. A case
study of influenza in waterfowl. Epidemics. 2009. [In press] (Doi: 10.1016/j.
epidem.2009.04.002).
13. Chowell G, Miller MA, Viboud C. Seasonal influenza in the United States,
France, and Australia: transmission and prospects for control. Epidemiol
Infect. 2008;136(6):852-64.
This article was published on 4 June 2009.
Citation style for this article: Nishiura H, Castillo-Chavez C, Safan M, Chowell G.
Transmission potential of the new influenza A(H1N1) virus and its age-specificity
in Japan. Euro Surveill. 2009;14(22):pii=19227. Available online: http://www.
eurosurveillance.org/ViewArticle.aspx?ArticleId=19227
4. Mills CE, Robins JM, Lipsitch M. Transmissibility of 1918 pandemic influenza.
Nature. 2004; 432(7019):904-6.
Figure 4
Estimates of the reproduction number for the epidemic of new influenza A(H1N1) virus infection in Japan
A) Estimated reproduction number R, based on the initial growth phase of the epidemic (i.e. first eight days)
B) Effective reproduction number Rt , as a function of time
A
B
6
9.0
8.0
Effective reproduction number
Reproduction number
5
4
3
2
1
Mean generation time (days)
7.0
2.5
6.0
1.9
5.0
1.3
4.0
3.0
2.0
1.0
0
3.4
3.7
4.0
23-May
3.1
21-May
2.8
19-May
2.5
17-May
2.2
15-May
1.9
13-May
1.6
11-May
1.3
9-May
0.0
1.0
Mean generation time (days)
Note:
A) Mean and variance of the generation time were 1.9 days and 0.8 days2 (given a coefficient of variation of 47%), and the sensitivity of R to different
mean generation times is examined. Coefficient of variation is kept constant when the mean generation time is varied.
B) R t > 1 indicates growth of cases at a given point of time, while R t < 1 indicates that the epidemic is in declining trend and may be under control. The
horizontal dashed line represents the threshold value, R t = 1. It should be noted that the dates of onset in Japan have yet to be refined, and the precision
of R t estimate may have been influenced by possible delay in diagnosis and reporting
4
E U ROSU R V E I L L A N C E Vol . 14 · I ss u e 22 · 4 J u n e 20 0 9 · w w w. e u ro s u rve i ll an c e . o rg